052182916X cambridge university press votes and violence electoral competition and ethnic riots in india jul 2004

P1: JZP052182916Xc01 CY386B/Wilkinson 0 521 82916 X April 2, 2004 17:36 Votes and Violence The Electoral Incentives for Ethnic Violence My central argument is that town-level electoral i

Trang 2

Votes and Violence

Why does ethnic violence break out in some places and not others? Moreimportant, why do some governments try to prevent antiminority riots whileothersdo nothing, or even actively encourage attacks? Thisbook answersthesequestions through a detailed study of Hindu-Muslim riots in India, as well ascase studies of Ireland, Malaysia, and Romania It shows how electoral incentives

at two levelsinteract to explain both where violence breaksout and, moreimportantly, why some states decide to prevent mass violence and others donot While developing thiselectoral incentivesmodel, the author showswhyseveral alternative explanations for ethnic violence – focusing on town-levelsocial and economic factors, the weak capacity of the Indian state, or India’salleged lack of “consociational power sharing” – cannot explain the observedvariation in Hindu-Muslim riots

Steven I Wilkinson is Assistant Professor of Political Science at Duke sity, where he has taught since 1999 He has traveled extensively in India sincehis ﬁrst visit there in 1989 He has been awarded fellowships from the HarvardAcademy for International and Area Studies, the Columbia University Society

Univer-of Fellowsin the Humanities, and the Harry Frank Guggenheim Foundation

Trang 3

P1: JZP/UKS P2: JZP/JZN P3: JZP/JZN QC:

052182916Xpre CY386B/Wilkinson 0 521 82916 X May 3, 2004 12:28

Trang 4

Cambridge Studies in Comparative Politics

General Editor

Margaret Levi University of Washington, Seattle

Assistant General Editor

Stephen Hanson University of Washington, Seattle

Associate Editors

Robert H Bates Harvard University

Peter Hall Harvard University

Peter Lange Duke University

Helen Milner Columbia University

FrancesRosenbluth Yale University

Susan Stokes University of Chicago

Sidney Tarrow Cornell University

Other Books in the Series

Stefano Bartolini, The Political Mobilization of the European Left,

1860–1980: The Class Cleavage

Mark Beissinger, Nationalist Mobilization and the Collapse of the Soviet State CarlesBoix, Democracy and Redistribution

CarlesBoix, Political Parties, Growth, and Equality: Conservative and Social Democratic Economic Strategies in the World Economy

Catherine Boone, Merchant Capital and the Roots of State Power in Senegal, 1930–1985

Catherine Boone, Political Topographies of the African State: Territorial Authority and Institutional Change

Michael Bratton and Nicolasvan de Walle, Democratic Experiments in Africa: Regime Transitions in Comparative Perspective

Valerie Bunce, Leaving Socialism and Leaving the State: The End of

Yugoslavia, the Soviet Union, and Czechoslovakia

Ruth BerinsCollier, Paths toward Democracy: The Working Class and Elites

in Western Europe and South America

Donatella della Porta, Social Movements, Political Violence, and the State Gerald Easter, Reconstructing the State: Personal Networks and Elite Identity Robert F Franzese, Macroeconomic Policies of Developed Democracies

Continues after the Index

Trang 5

Trang 6

ELECTORAL COMPETITION AND ETHNIC RIOTS IN INDIA

STEVEN I WILKINSON

Duke University

Trang 7

cambridge university press

Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São PauloCambridge University Press

The Edinburgh Building, Cambridge cb2 2ru, UK

First published in print format

isbn-13 978-0-521-82916-8

isbn-13 978-0-511-21709-8

2004

Information on this title: www.cambridge.org/9780521829168

This publication is in copyright Subject to statutory exception and to the provision ofrelevant collective licensing agreements, no reproduction of any part may take placewithout the written permission of Cambridge University Press

isbn-10 0-511-21709-9

isbn-10 0-521-82916-x

Cambridge University Press has no responsibility for the persistence or accuracy of urlsfor external or third-party internet websites referred to in this publication, and does notguarantee that any content on such websites is, or will remain, accurate or appropriate

Published in the United States of America by Cambridge University Press, New Yorkwww.cambridge.org

hardback

eBook (NetLibrary)eBook (NetLibrary)hardback

Trang 9

Figures

1.1 The relationship between party competition and a

state’s response to antiminority polarization and

1.3 State variation in deaths in Hindu-Muslim riots,

1.4 State variation in number of Hindu-Muslim riots,

2.2 Reported precipitating eventsand deathsduring the

5.1 The theoretical relationship between party competition

and a state’s response to antiminority polarization and

5.2 Predicted effect of party fractionalization on communal

riots1525.3 Reported precipitating eventsand deathsduring the

February–April 2002 communal violence and patterns

6.1 The institutional origins of state-level differences

7.1 The relationship between party competition and

a state’s response to antiminority polarization

and violence: Non-Indian examples206

Trang 10

2.1 Electoral competition and occurrence of riots

3.3 Frequency of police transfers in major states,

3.4 Police strength, judicial capacity, and riots in

the 1990s883.5 Arrests, prosecutions, and convictions after

communal riots893.6 State transfer rates and Hindu-Muslim riots,

Trang 11

5.3 Electoral competition and communal riots

5.4 Party competition and riot prevention,

6.1 Do state-level differences in ethnic heterogeneity

7.1 Election resultsand ethnic cleavagesin Malaysian

Trang 12

This study owes a great deal to my advisor at MIT, the late and much missedMyron Weiner, who along with Donald Horowitz and Stephen Van Everaurged me to test ideas I had developed on the basis of a study of UttarPradesh in a much wider comparative study

I have been fortunate to have the opportunity to present large portions

of the book to very smart members of two cross-disciplinary groups, whoseconstructive comments have helped me to identify areas where the argu-ment or evidence needed more work Membersof the Laboratory in Com-parative Ethnic Processes (LiCEP) have read several drafts of the mainargument over several years, though they cannot be blamed, of course, for

my inability or unwillingness to incorporate all their suggestions I thank allthe current and former membersof thisterriﬁc group for their help: ArunAgrawal, Bob Bates, Kanchan Chandra, Christian Davenport, Jim Fearon,Karen Ferree, Elise Giuliano, Michael Hechter, Macartan Humphries,Stathis Kalyvas, Nelson Kasﬁr, Pauline Jones Luong, David Laitin, IanLustick, Dan Posner, Nicholas Sambanis, Smita Singh, Pieter van Houten,and Ashutosh Varshney Several members of the group – Kanchan Chandra,Jim Fearon, Dan Posner, Nic Sambanis, and Elisabeth Wood – gave me

so many additional detailed comments and suggestions that I should givethem my special thanks The second group that has read my work, therecently formed Network on South Asian Politics and Political Economy(Netsappe), commented on an earlier version of my main chapters at aconference at the University of Michigan in July 2002, and participantsprovided feedback that reﬂected both the theoretical and area expertise ofthe group I thank all the membersof Netsappe for giving me extremelydetailed and insightful comments on the two chapters I presented Special

Trang 13

me clarify my central argument about the relationship between party petition and violence at low levelsof party competition Were it not for ourconversationstogether, the key diagramsin which I work out the relation-ship between party fractionalization and government response to anti-minority violence would be one “branch” short Herbert and John Aldrichalso encouraged me to present several key chapters to the party politicscourse they coteach, from which I received additional valuable feedback.Otherswhose help went above and beyond the call of duty include JohnTransue, Scott de Marchi, and Meg McKean Lastly, I should like to thankMike Munger, not for hisfriendly nagging about whether I would ever getthis book ﬁnished (though that too was appreciated, at least most of thetime), but for his interest in my work and steady encouragement over thepast ﬁve years.

com-In com-India, many people have helped me at variouspointsin my research

In Delhi, Sunil and Anjali Kumar and their family have given me theirfriendship and encouragement – and often food and shelter as well – since

my ﬁrst visit to India in 1989 They make my trips to India seem like cominghome Otherswho provided hospitality and encouragement include Tejbirand Mala Singh in Delhi, Gyan and Jayati Chaturvedi in Agra, and S K.Gupta in Varanasi Many police ofﬁcers and civil servants offered to sharetheir expertise with me, and I would especially like to note the help I receivedfrom the late Ashok Priyadarshi, N S Saksena, and A K Dass

Otherswho have helped along the way include Neil Carlson, who was

a huge help in solving data management and aggregation issues, CharlesFranklin, who offered advice on the statistical model to use, and David Clineand Carrie Young, whose work on the regional database was invaluable.Paul Brass, whose own work on ethnic conﬂict is central to the ﬁeld, readand commented on approximately half the manuscript I am sure that westill do not agree, but I thank him for taking the time to offer his incisiveand constructive comments Devesh Kapur also deserves thanks for his

Trang 14

my profound thanks.

Partsof Chapter 4 were originally published as“India, Consociational

Theory and Ethnic Violence,” Asian Survey 40, no 5 (October 2000),

pp 767–91

Trang 15

Trang 16

The Electoral Incentives for Ethnic Violence

In the 1960s Richard Nixon, reflecting on race riots in America, tried to fine the difference between riots and other types of violent conflict “Riots,”

de-he said, “are spontaneous Wars require advance planning.”1My argument

in this book, by contrast, is that ethnic riots, far from being relatively taneous eruptions of anger, are often planned by politicians for a clear elec-toral purpose They are best thought of as a solution to the problem of how

spon-to change the salience of ethnic issues and identities among the elecspon-torate inorder to build a winning political coalition Unpleasant as this ﬁnding may

be, political competition can lead to peace as well as violence, and I identifythe broad electoral conditions under which politicians will prevent ethnicpolarization and ethnic violence rather than incite it I demonstrate, usingsystematic data on Hindu-Muslim riots in India, that electoral incentives

at two levels – the local constituency level and the level of government thatcontrols the police – interact to determine both where and when ethnicviolence against minorities will occur, and, more important, whether thestate will choose to intervene to stop it

Pointing out that there is a relationship between political competitionand ethnic violence is not in itself new Ethnic violence has often beenportrayed as the outcome of a rational, if deplorable, strategy used bypolitical elites to win and hold power Bates, for example, argued twodecades ago that in Africa, “electoral competition arouses ethnic conﬂict.”2

1 Richard M Nixon, “The War in Our Cities,” address before the National Association of

Manufacturers, New York City, December 8, 1967, quoted in James J Kilpatrick, Evening Star (Washington, D.C.), December 26, 1967, p A13.

2 Robert H Bates, “Modernization, Ethnic Competition and the Rationality of Politics in

Contemporary Africa,” in Donald Rothchild and Victor Olorunsola, eds., State versus Ethnic Claims: African Policy Dilemmas (Boulder: Westview Press, 1983), p 161.

Trang 17

P1: JZP

052182916Xc01 CY386B/Wilkinson 0 521 82916 X April 2, 2004 17:36

And many scholars have since blamed the upsurge of ethnic violence inEastern Europe in the 1990s on the strategies of ex-Communist politi-cians like Miloˇsevi´c who used ethnic nationalism to distract attention fromtheir own past sins and their countries’ present economic and social prob-

basis of a worldwide survey of ethnic violence in the 1990s, that ethnic riots

and pogroms are usually caused by political elites who “play on existing

communal tensions to entrench [their] own power or advance a politicalagenda.”4

There are, however, at least three reasons why I ﬁnd most “instrumental”political explanations for violence to be unsatisfying First, because scholars

who study ethnic violence generally look at political elites who have incited

ethnic violence, they offer us little insight into why some politicians seem to

do exactly the opposite and use their political capital and control of the state

to prevent ethnic conﬂict Why, for example, did President Boigny of C ˆote d’Ivoire respond to attacks on traders from the Mauritanianminority in Abidjan in 1981 by sending police to protect Mauritaniansand then going on national radio to praise Ivoirians who had guarded thetraders’ property while they were under police protection?5 Why morerecently in India was Chief Minister Narendra Modi of Gujarat so weak inresponding to large-scale anti-Muslim violence in his state, whereas otherchief ministers such as Chandrababu Naidu in Andhra Pradesh or DigvijaySingh in Madhya Pradesh were successful in preventing riots in their states?6

Houphouet-Second, many political explanations for ethnic violence fail to account for

3 Claus Offe, “Strong Causes, Weak Cures: Some Preliminary Notes on the Intransigence

of Ethnic Politics,” East European Constitutional Review 1, no 1 (1992), pp 21–23; Tom Gallagher, Romania after Ceausescu: The Politics of Intolerance (Edinburgh: Edinburgh Univer-

sity Press, 1995), pp 3–5 For an examination of the role of elites in preventing compromise and exacerbating the security dilemma, see Stuart Kaufman, “The Irresistible Force and the

Imperceptible Object: The Yugoslav Breakup and Western Policy,” Security Studies 4, no 2

(1994–95), p 282.

4Human Rights Watch, Slaughter among Neighbors: The Political Origins of Communal Violence

(New Haven: Yale University Press, 1995), pp 2, 7, 65–66 (emphasis added).

5FBIS (West Africa), April 21–22, 1980, p T4; Tanzanian Daily News, March 12, 1981; West Africa, September 30, 1985, p 2064; Le Monde, September 6, 1985; Economist Information Unit Country Report #1: Cˆote d’Ivoire 1992 (London: Economist Information Unit, 1992),

p 12.

6Steven I Wilkinson, “Putting Gujarat in Perspective,” Economic and Political Weekly

(Mumbai), April 27, 2002, pp 1579–83 For details of the Gujarat government response

to the riots, see “‘We Have No Orders to Save You’: State Participation and Complicity in

Communal Violence in Gujarat,” Human Rights Watch 14, no 3 (C) (2002).

Trang 18

Electoral Incentives for Ethnic Violence

the variation in patterns of violence within states In part because elitetheories of ethnic violence focus on the strategies and actions of national-level political leaders such as Franjo Tudman and Slobodan Miloˇsevi´c in−former Yugoslavia or Daniel Arap Moi in Kenya, they cannot explain why,within a state, violence breaks out in some towns and regions but not in manyothers Why, for example, when the 1969 riots in Malaysia were allegedlyabout national-level political issues, did riots break out in Kuala Lumpurand elsewhere in Selangor state but not in the states of Penang, Johore,and Kedah?7 Why in India did riots over the “national” issue of the BabriMasjid–Ram Janambhoomi site in 1989–92 take place in some towns andstates but not in others? Third, the role of political incentives in fomentingviolence is generally “proven” from the simple fact that ethnic violence hasbroken out and that some politician gained from the outbreak; seldom arepolitical incentives independently shown to exist and to be responsible forthe riots

My aim in this book is to understand why Hindu-Muslim violence takesplace in contemporary India, which necessarily involves addressing threegeneral problems in the instrumentalist literature on ethnic violence.8First,

I want to account for interstate and town-level variation in ethnic violence inIndia: why do apparently similar towns and states have such different levels

of violence? Second, when dealing with the role of the political incentivesfor ethnic violence, I want to understand the conditions under which thepoliticians who control the police and army have an incentive both to fo-ment and to prevent ethnic violence Third, I want to demonstrate thatthe political incentives I identify as important actually work in the way Isuggest, by tracing through individual cases where politicians fomented orrestrained violence

7 William Crego Parker, “Cultures in Stress: The Malaysian Crisis of 1969 and Its Cultural Roots” (Ph.D dissertation, MIT, 1979), 1:183.

8 I treat Hindus and Muslims as “ethnic groups” in the sense that Weber deﬁnes them,

as having a “subjective belief in their common descent because of similarities of cal type or of customs or both, or because of memories of colonization and migration.”

physi-Max Weber, Economy and Society: An Outline of Interpretive Sociology, vol 1, ed Guenther

Roth and Claus Wittich (Berkeley: University of California Press, 1978), p 389 For ers who integrate a discussion of Hindu-Muslim violence into their general theories of

oth-ethnic conﬂict, see Donald L Horowitz, Ethnic Groups in Conﬂict (Berkeley: University of California Press, 1985), pp 50–51; John Breuilly, Nationalism and the State (Chicago: Uni- versity of Chicago Press, 1994), pp 206–15; Ashish Nandy et al., Creating a Nationality: The Ramjanmabhumi Movement and Fear of the Self (Delhi: Oxford University Press, 1995),

p vi.

Trang 19

P1: JZP

The Electoral Incentives for Ethnic Violence

My central argument is that town-level electoral incentives account forwhere Hindu-Muslim violence breaks out and that state-level electoral in-centives account for where and when state governments use their policeforces to prevent riots We can show that these town- and state-level elec-toral incentives remain important even when we control for socioeconomicfactors, local patterns of ethnic diversity, and towns’ and states’ previouslevels of Hindu-Muslim conﬂict

At the local level I begin with the constructivist insight that individualshave many ethnic and nonethnic identities with which they might identifypolitically.9The challenge for politicians is to try to ensure that the iden-tity that favors their party is the one that is most salient in the minds of amajority of voters – or a plurality of voters in a single-member district sys-tem – in the run-up to an election I suggest that parties that represent eliteswithin ethnic groups will often – especially in the most competitive seats –use polarizing antiminority events in an effort to encourage members oftheir wider ethnic category to identify with their party and the “majority”identity rather than a party that is identiﬁed with economic redistribution

or some ideological agenda These antiminority events, such as ing a dispute over an Orange Lodge procession route through a Catholicneighborhood in Ireland, or carrying out a controversial march around

provok-a disputed Hindu temple or Muslim mosque site in Indiprovok-a, provok-are designed

to spark a minority countermobilization (preferably a violent mobilization that can be portrayed as threatening to the majority) that willpolarize the majority ethnic group behind the political party that has thestrongest antiminority identity.10When mobilized ethnic groups confronteach other, each convinced that the other is threatening, ethnic violence isthe probable outcome

counter-Local electoral incentives are very important in predicting where lence will break out, though as I discuss in Chapter 2 they are not theonly local-level factor that precipitates or constrains ethnic riots Ulti-mately, however, there is a much more important question than that of

vio-9 For a survey of how “constructivist” research has affected the study of ethnic conﬂict, see the special issue of the American Political Science Association’s comparative politics

newsletter devoted to “Cumulative Findings in the Study of Ethnic Politics,” APSA – CP

Newsletter 12, no 1 (2001), pp 7–22.

10 An important enabling condition here is the presence of some preexisting antiminority sentiment among members of the ethnic majority.

Trang 20

the local incentives for violence: the response of the level of governmentthat controls the police or army In virtually all the empirical cases I haveexamined, whether violence is bloody or ends quickly depends not onthe local factors that caused violence to break out but primarily on thewill and capacity of the government that controls the forces of law andorder

Abundant comparative evidence shows that large-scale ethnic riotingdoes not take place where a state’s army or police force is ordered to stop

it using all means necessary The massacres of Chinese in Indonesia inthe 1960s, for instance, could not have taken place without the Indonesianarmy’s approval: “In most regions,” reports Robert Cribb, “responsibilityfor the killings was shared between army units and civilian vigilante gangs

In some cases the army took direct part in the killings; often, however, theysimply supplied weapons, rudimentary training and strong encouragement

to the civilian gangs who carried out the bulk of the killings.”11 nority riots in Jacksonian America were also facilitated by the reluctance

Antimi-of local militias and sheriffs to intervene to protect unpopular minorities.12

And recent ethnic massacres in Bosnia, Rwanda, and Burundi were wise possible only because the local police forces and armies refused tointervene against or even directly participated in the violence.13Finally, theworst partition massacres in India in 1946–47took place in those provinces –Bengal, Punjab, and Bihar – in which the elected local governments, eachcontrolled by the majority ethnic group, made it plain at various times thatthey would not intervene against “their” community to protect the ethnicminority from attack In Bihar, for example, after anti-Muslim riots brokeout in October 1946 the province’s Hindu premier refused to allow Britishtroops to ﬁre on Hindu rioters, ignored Congress leaders’ complicity in theriots, held no ofﬁcial inquiry, and made only a few token arrests of thosewho had participated in anti-Muslim pogroms that killed 7,000 to 8,000people.14

like-11 Robert Cribb, “Problems in the Historiography of the Killings in Indonesia,” in Cribb, ed.,

The Indonesian Killings, 1965–66: Studies from Java and Bali (Melbourne: Centre for South

East Asian Studies, Monash University, 1990), p 3.

12Michael Feldberg, The Turbulent Era: Riot and Disorder in Jacksonian America (New York:

Oxford University Press, 1980), pp 28, 111.

13See, e.g., Ren´e Lemarchand, Burundi: Ethnic Conﬂict and Genocide (New York: Woodrow

Wilson Center Press/Cambridge University Press, 1996), pp 96–100.

14Vinita Damodaran, Broken Promises: Popular Protest, Indian Nationalism and the Congress Party

in Bihar, 1935–1946 (Delhi: Oxford University Press, 1992), pp 354–56.

Trang 21

P1: JZP

Three or more parties

(3.5+ ENPV)

1 Most Indian states in 2002:

e.g Kerala, Bihar, Orissa

B i

Government does not rely on minority votes

1 One Indian state in 2002: Gujarat

2 Romanian national govt 1990

3 State & local governments in US South 1877-1960s

4 Irish local governments in early 19 th

1 Three Indian states in 2002:

Andhra Pradesh, Madhya Pradesh, Rajasthan

2 USA national-level post-1948

Government will not prevent riots

Government prevents riots

A

Figure 1.1 The relationship between party competition and a state’s response to

antiminority polarization and violence: Indian and non-Indian examples (ENVP=effective number of parties)

If the response of the state is the prime factor in determining whether nic violence breaks out, then what determines whether the state will protectminorities? My central argument is that democratic states protect minoritieswhen it is in their governments’ electoral interest to do so (see Figure 1.1).Speciﬁcally, politicians in government will increase the supply of protection

eth-to minorities when either of two conditions applies: when minorities are animportant part of their party’s current support base, or the support base ofone of their coalition partners in a coalition government; or when the overall

Trang 22

electoral system in a state is so competitive – in terms of the effective number

of parties – that there is therefore a high probability that the governing partywill have to negotiate or form coalitions with minority supported parties inthe future, despite its own preferences.15The necessity to engage in whatHorowitz calls “vote-pooling” in order to win elections and maintain coali-tions is what forces politicians to moderate their demands and offer protec-tion to minorities “The prospect of vote pooling with proﬁt,” as he pointsout, “is the key to making parties moderate and producing coalition withcompromise in severely divided societies.”16In India, vote pooling moder-ates even the behavior of nationalist parties that have no minority support,

as long as these parties are forced to form coalitions with parties that do rely

on minority votes On the other hand, politicians in government will restrictthe supply of security to minorities if they have no minority support and theoverall levels of party competition in a state are so low that the likelihood

of having to seek the support of minority-supported parties in the future isvery low

In addition to these three competitive situations, Figure 1.1, lists theIndian states in each category (as of February 2002) Most Indian states to-

day fall into category A, where the presence of high levels of party

competi-tion (3.5–8 effective parties, using the effective number of parties or ENPVmeasure) forces politicians to provide security to minorities because to dootherwise would be to destroy present-day coalitions as well as future coali-tional possibilities.17 A handful of Indian states falls into category B, with

bipolar party competition (which amounts to 2–3.5 effective parties using

15 The formula for the effective number of parties is ENPV= 1/vi2, where v i is the vote share of the ith party This widely used measure weights parties with a higher vote share

more heavily than those parties with a very low vote share, thus providing a better measure

of the “real” level of party competition than if we were to simply count the total number

of parties competing in a state.

16Donald L Horowitz, A Democratic South Africa: Constitutional Engineering in a Divided Society (Berkeley: University of California Press, 1991), pp 177–83 (quotation from

p 177).

17 The effective number of parties (votes) or ENPV is a measure that places higher weight

on parties with high vote shares than parties with very low vote shares, thus providing a much better measure of the “true” level of party competition than if we were simply to count the total number of parties competing in a state election For example if we were simply to count the total number of parties competing in the Gujarat and Madhya Pradesh state elections of 1998 (17parties and 41 parties, respectively), we would have a misleading impression of the true level of party competition in these states, because both states in 1998 were in fact two horse races between the BJP and the Congress, with the BJP and Congress obtaining 93.4% of the total votes between them in Gujarat and 80% in Madhya Pradesh The effective number of votes measure (ENPV) of 2.97parties for Gujarat and 3.09 parties

Trang 23

P1: JZP

the ENPV measure) In 2002 there were four large Indian states with suchbipolar patterns of party competition: Gujarat, Madhya Pradesh, AndhraPradesh, and Rajasthan Three of these states – Andhra Pradesh, Madhya

Pradesh, and Rajasthan – fall into subcategory B i, in which the party inpower in the state relied heavily on a multiethnic supportbase that includessubstantial or overwhelming Muslim support Only in Gujarat in 2002 did

we have the worst-case scenario (subcategory B ii) where there were bothlowlevels of party competition in the state (2.97effective parties) and a gov-ernment in power, the Bharatiya Janata Party (BJP), that did not have anyminority support base and therefore had no incentive to protect Muslims.The reaction of state governments to violence in 2002 is predicted almostperfectly by their degrees of party competition and minority support, as Idiscuss in Chapter 5

The basic electoral incentives model presented here can easily be tended to account for patterns of government riot-prevention in othermultiethnic democracies as well (see Chapter 7).18 In looking at patterns

ex-of state riot prevention in the U.S South, for example, the key explanatoryfactor that explains greater federal government willingness to intervene toprotect African Americans after World War II was the fact that black voterswho had emigrated from the South between 1910 and 1950 became a vitalconstituency for the Democratic Party in several important swing states in

the north, such as Michigan and Illinois This shift (from category B iito

category B i in Figure 1.1) prompted northern Democratic leaders ﬁnally

to intervene in the South to protect the civil rights of African Americans.19

for Madhya Pradesh represents this true level of competition much better than counting the total number of parties.

18 Although the argument I develop in this book applies to democratic governments, in ciple there is no reason why it could not also be extended to explain the conditions under which authoritarian governments will prevent antiminority violence Authoritarian regimes need not be concerned about voters, but they still have to be concerned about constituencies that can offer ﬁnancial, political, and military support If an ethnic minority is well placed to offer such support to an authoritarian regime, then we would expect the regime

prin-to protect the minority even if it is very unpopular with the majority of the population.

In Indonesia, for example, the Chinese minority did well under Suharto because it offered ﬁnancial support, but the Chinese have done less well in a democracy.

19 In India the day-to-day responsibility for law and order rests with the states, not with local or federal governments Therefore explaining where and when antiminority violence breaks out and whether it is suppressed by the state in India is explicable by looking at electoral incentives at two levels In cases where, as in the United States, local, county, state, and national authorities all have shared authority over local law enforcement, then

Trang 24

To give another example: in Ireland in the 19th century the high levels

of Protestant-Catholic violence in Belfast in the early 1860s comparedwith that in other cities in Ireland can be explained by the fact that thepolice force in Belfast, unlike elsewhere in the country, was locally con-trolled by a Protestant-majority town council that did not rely on Catholicvotes and therefore had no electoral incentive to intervene to protect

Catholics from Protestants (situation B ii) Only once the control of localpolicing was taken away from the Belfast council in 1865 and transferred

to a national administration that was determined to prevent Catholic violence do we see a signiﬁcant increase in the state’s degree ofriot prevention

Protestant-Testing the Electoral Incentives Explanation

One general problem in testing theories of ethnic violence is that in mostcases we lack systematic data on ethnic riots or their likely economic, socialand political causes.20There is, for example, no equivalent for intranationalethnic violence of the massive “Correlates of War” project in internationalrelations, which collects data on all international violence from 1816 to

1980.21 In the past decade several scholars have tried to collect detaileddata on ethnic violence in the former Soviet Union, where Western secu-rity interests, and hence foundation research funds, are substantial.22Butpolitical scientists have not yet matched the efforts of their colleagues inhistory in collecting basic information about each country’s internal pattern

the model outlined here can simply be extended to incorporate electoral incentives and power asymmetries across different levels of governments.

20 The United States is the obvious exception to this general statement I have been able to identify only one study on ethnic violence in the developing world that collects systematic

intranational data: Remi Anifowose, Violence and Politics in Nigeria: The Tiv and Yoruba Experience (New York: Nok Publishers, 1982).

21 For a review of the research the Correlates of War project inspired, see John A Vasquez,

“The Steps to War: Towards a Scientiﬁc Explanation of Correlates of War Findings,” World Politics 40, no 1 (1988), pp 109–45.

22 Marc Beissinger at the University of Wisconsin has collected information on all reported

“nationalist mobilization” and violence in the Former Soviet Union from 1987to 1991 See Beissinger, “How Nationalisms Spread: Eastern Europe Adrift the Tides and Cycles

of Nationalist Contention,” Social Research 63, no 1 (1996), pp 97–146 Ian Bremmer and

Ray Taras provide a “Chronology of Ethnic Unrest in the USSR, 1985–92,” in their edited

volume Nations and Politics in the Soviet Successor States (Cambridge: Cambridge University

Press, 1993), pp 539–49.

Trang 25

P1: JZP

of ethnic riots before putting forward theories to explain why they occur inone place and not another.23

A few pioneering collaborative projects have collected aggregate tics on the largest incidents of ethnic violence reported by the Western me-dia.24But for my purposes, these surveys underreport small and nondeadlyethnic riots, which account for the majority of incidents in most countries

statis-In statis-India, for example, press data suggest that most Hindu-Muslim riotslead to no deaths and 80% of those riots in which deaths do occur aremuch smaller in size (1–9 deaths) than would typically prompt a report inthe international news media Moreover, the aggregate data provided bysuch studies as the Minorities at Risk project, though good for interstatecomparisons, do not provide the detailed town-by-town information onviolence that would allow us to test many of the leading microtheories ofethnic conﬂict

In this book I test my electoral explanation argument for ethnic riotsusing state- and town-level data on Hindu-Muslim riots in India over thepast ﬁve decades.25To address the lack of good data on town- and state-levelethnic violence in India, I utilize a new dataset on Hindu-Muslim riots inIndia, jointly collected by myself and Ashutosh Varshney, now at the Uni-versity of Michigan The 2,000 riots in the database cover the years 1950–

95 When combined with a separate database I collected independently

23 For historical research in which systematic data collection on riots plays a major role in

theory testing, see Manfred Gailus, “Food Riots in Germany in the Late 1840s,” Past and Present 145 (1994), pp 157–93; James W Tong, Disorder under Heaven: Collective Violence

in the Ming Dynasty (Stanford: Stanford University Press, 1991); John Bohstedt, “Gender, Household and Community Politics: Women in English Riots, 1790–1810,” Past and Present

120 (1988), pp 88–122; Frank Neal, Sectarian Violence: The Liverpool Experience, 1819–1914

(Manchester: Manchester University Press, 1988).

24 Notably the Minorities at Risk Project at the University of Maryland, which covers

c 300 ethnic groups See Ted Robert Gurr and Barbara Harff, Ethnic Conﬂict in World Politics (Boulder: Westview Press, 1994) For details, see the project’s web site at

<http://www.bsos.umd.edu/cidcm/mar/indmus.htm>.

25 Donald L Horowitz deﬁnes a “deadly ethnic riot,” as “an intense, sudden, though not necessarily wholly unplanned, lethal attack by civilian members of one ethnic group on civilian members of another ethnic group, the victims chosen because of their group mem-

bership.” Horowitz, The Deadly Ethnic Riot (Berkeley: University of California Press, 2001),

p 1 I define “Hindu-Muslim riots” in essentially the same way in this book, dropping only the “lethal” requirement in Horowitz’s definition of “deadly riots.” Hindu-Muslim riots often lead to deaths and injuries, but sometimes they do not For alternative definitions, see

Susan Olzak, The Dynamics of Ethnic Competition and Conﬂict (Stanford: Stanford University

Press, 1992), pp 233–34; Richard D Lambert, “Hindu-Muslim Riots” (Ph.D dissertation, University of Pennsylvania, 1951), p 15.

Trang 26

for the years 1900–49, the dataset represents the most comprehensive isting source on Hindu-Muslim violence (for full details, together with

ex-a protocol explex-aining how events were coded, see Appendixes A ex-and B).Collecting these data on Hindu-Muslim riots involved reading through ev-ery single issue of India’s newspaper of record from 1950 to 1995, as well

as (for my 1900–49 data) hundreds of reports in other newspapers, cial government reports, and archives in India, England, and the UnitedStates Because the resulting data are town-level as well as state-level, andextend back more than a century (unlike Government of India aggregateﬁgures on communal violence, which have only been published since 1954),they allow me to test theories of Hindu-Muslim violence much more com-pletely than has been done before, which should increase conﬁdence in myconclusions.26

ofﬁ-In addition to this effort to gather material on Hindu-Muslim riots, Ialso spent several years gathering town- and state-level data in India andfrom Indian government documents with which to operationalize and testthe main theories of ethnic violence For example, to test institutional decaytheories, which argue that a decline in the state’s bureaucratic and coercivecapacity leads to ethnic violence, I gathered data on politically motivatedtransfer rates, the changing ethnic and caste balance of the police and ad-ministration, and statistics on corruption To test economic theories thatfocus on town-level Hindu-Muslim economic competition, I combinedcensus data on employment with case studies, surveys, and governmentdirectories on particular handicrafts to develop a dummy variable that in-dicates whether, according to the theory, any particular town is likely tosuffer from communal violence.27And to test ecological theories that arguethat the Hindu-Muslim population balance or presence of Hindu refugeescauses riots, I used a mix of census data, poverty data, and World Bank datathat I collected for all major Indian states

26 For examples of the way in which post-1954 government data are used by scholars, see

Paul Brass, The Politics of India since Independence (Cambridge: Cambridge University Press, 1990), p 199; Atul Kohli, Democracy and Discontent: India’s Growing Crisis of Governability

(Cambridge: Cambridge University Press, 1990), p 7; Lloyd I Rudolph and Suzanne

Hoeber Rudolph, In Pursuit of Lakshmi: The Political Economy of the Indian State (Chicago:

University of Chicago Press, 1987), pp 226–27.

27The main all-India survey I use is S Vijayagopalan, Economic Status of Handicraft Artisans

(New Delhi: National Council for Applied Economic Research, 1993) The Uttar Pradesh government also publishes directories that allow us to establish religious breakdowns for

wholesalers and self-employed artisans See, e.g., Uttara Pradesha Vyapar Protsahan hikaran (Udhyog Nirdeshalaya: Kanpur, 1994).

Trang 27

Riots per 10 Million

Deaths in Hindu-Muslim Riots per 10 million Riots per 10 million

Figure 1.2 Hindu-Muslim riots since independence (data from Varshney and

Wilkinson)

The Importance of Understanding Hindu-Muslim Violence

For those concerned about the welfare of the world’s most populous racy, understanding the causes of Hindu-Muslim riots is of more than justtheoretical importance Hindu-Muslim riots threaten the stability of theIndian state, its economic development, and the country’s delicate inter-national relations with its Muslim neighbors, especially its nuclear-armedrival Pakistan Since the 1950s, as we can see in Figure 1.2, the number andgravity of Hindu-Muslim riots in India has grown to alarming proportions,reaching a dangerous peak in 1992–93, when nationwide riots broke out af-ter the destruction by Hindu militants of the Babri mosque in the northernIndian town of Ayodhya Since 1992 there has also been one further majoroutbreak of mass rioting, in the western state of Gujarat in 2002, in which

democ-an estimated 850 to 2,000 people were murdered.28

By some measures the numbers involved may not seem large The proximately 10,000 deaths and 30,000 injuries that have occurred in re-ported Hindu-Muslim riots since 1950 are, after all, only a fraction of the

ap-28 “We Have No Orders to Save You,” p 4.

Trang 28

60,000 people who die on India’s chaotic, congested roads each year, andthe annual rate of deaths from Hindu-Muslim riots is much lower than that

of the number of women murdered in so-called “dowry deaths” (3,000–4,000).29India’s per-capita death rate from Hindu-Muslim riots is also lowwhen compared with that in some of the world’s other well-known ethnicconﬂicts For example, deaths in Northern Ireland since 1969 run at 50times the per-capita rate in India due to Hindu-Muslim violence.30

But the importance of the Hindu-Muslim divide lies in more than justthe number of people who have died in riots since independence The divide

is also important because the Hindu-Muslim cleavage has split the Indianstate apart once already and has the potential to do so again An estimated200,000 people were murdered and 13 million forced to migrate from theirhomes in 1946–48 when India was partitioned into Muslim and Hindu ma-jority states.31Because Hindus and Muslims live side by side throughout thelength and breadth of India, this cleavage poses a potentially much more se-rious threat to the country than separatist conﬂicts in the North and North-east, which have so far claimed a greater number of lives.32This is especially

so because Hindu-Muslim violence affects some states at some times somuch more than others As I show in Figures 1.3 and 1.4, which report data

on Hindu-Muslim riots after the 1977 emergency, states such as Gujaratand Maharashtra have had, even allowing for population, considerably high-

er average monthly levels of riots and deaths over the past three decades.33

Hindu-Muslim riots also have damaging, though often ignored, effects

on India’s economic development, and these effects again are concentrated

29 In 1989, for example, when the Ayodhya agitation was nearing its height, 521 people died

in communal riots compared to 3,894 women who were murdered over dowry Annexure

no 117, Rajya Sabha Debates, Appendix 155, August 7–September 7, 1990, pp 558–60 This

ofﬁcial rate of dowry deaths is of course widely recognized to be a gross underestimate.

30 According to 1995 Royal Ulster Constabulary (RUC) ﬁgures, 3,462 people have died in

the Northern Ireland conﬂict out a population of c 1.5 million Mervyn T Love, Peace Building through Reconciliation in Northern Ireland (Avebury: Aldershot, 1995), p 38.

31My estimate of deaths comes from Penderel Moon, Divide and Quit (London: Chatto and

Windus, 1961), p 269 Moon gives a clear explanation of how he arrived at this ﬁgure Scholarly and journalistic estimates that claim a million or more deaths are common but unsubstantiated Keller for instance quotes a ﬁgure of “up to 1 million” dead in communal

rioting Stephen L Keller, Uprooting and Social Change: The Role of Refugees in Development

(Delhi: Manohar Book Service, 1975), p 17.

32Horowitz, Ethnic Groups in Conﬂict, p 37

33 Interestingly, as we can see from Figures 1.3 and 1.4, riots seem to be much more evenly spread than casualties across states We will try to explain in subsequent chapters why, even though riots break out across India, they only seem to lead to large numbers of deaths in some states.

Trang 29

P1: JZP

Figure 1.3 State variation in deaths in Hindu-Muslim riots, 1977–1995: Monthly

average per 10 million inhabitants (based on data collected by Varshney and

Wilkinson from Times of India reports)

in certain states.34The Hindu-Muslim riots of January 1993, for example,cost the city of Mumbai (Bombay) alone an estimated Rs 9,000 crores ($3.6billion) in lost production, sales, tax revenues, property losses, and exportsand reportedly forced one industry, synthetic textiles, to at least temporarilyabandon Mumbai altogether.35Industries in which Muslims account for adisproportionately large share of the work force, such as leather, jewelry,

34 “Mosque Demolition: Consequences for Reform,” Economic Times (Bombay), December 10,

1992.

35 The Mumbai-based Noorani family, the owner of Zodiac clothing, temporarily ﬂed the city and has since directed its new investments outside Maharashtra, mainly in Bangalore Many

Indian statistics are given in units of a crore (ten million) or a lakh (hundred thousand) The

ﬁgure on total losses is from the business consultancy Tata Services, reported in Ashgar

Trang 30

Figure 1.4 State variation in the number of Hindu-Muslim riots, 1977–1995:

Monthly average per 10 million inhabitants (based on data collected by Varshney

and Wilkinson from Times of India reports)

bakeries, and textiles, were particularly hard hit.36In Mumbai’s ready-madegarment industry, for instance, where Muslims from the northern states ofUttar Pradesh and Bihar are employed in hand and machine embroidery, the

1993 migration of Muslims back to their towns and villages cost turers more than $3 million a day in lost production.37The Muslim exodusfrom Mumbai, by drying up remittances, further impoverished the econo-mies in the migrants’ home districts in Uttar Pradesh, Bihar, and Bengal

manufac-Ali Engineer, “Bombay Riots: Second Phase,” Economic and Political Weekly, March 20–27,

Trang 31

P1: JZP

Hindu-Muslim riots also endanger India’s international security and thesecurity of Hindus living outside India Every Hindu-Muslim riot increasestensions between Pakistan and India, South Asia’s two nuclear powers.38

Since the 1950s large anti-Muslim riots in India have often sparked for-tat violence against Hindu minorities in Pakistan and Bangladesh InDecember 1992 and January 1993, for example, anti-Muslim riots in Indiawere swiftly followed by serious anti-Hindu riots in Karachi, Lahore, andDhaka The mass migration of South Asians to other countries and thespread of global news media have also increased the likelihood that riots inIndia will lead to violence against Hindus far from India’s borders The 1992Hindu-Muslim riots had repercussions as far away as Dubai, Thailand, andBritain (where Muslim mobs in Bradford and other northern English citiesattacked Hindu temples).39

tit-Plan of the Book

I begin in Chapter 2 by examining the town-level causes of Hindu-Muslimriots and the broader question of intrastate variation in ethnic violence.Using systematic town-level data on riots and socioeconomic variables fromIndia’s most populous state, Uttar Pradesh, I show that the probability

of whether a town will have a Hindu-Muslim riot is highly related to itslevel of electoral competition, even once we hold factors such as a town’sdemographic balance or its past record of Hindu-Muslim violence constant.Towns with a close electoral race are considerably more likely to have aHindu-Muslim riot than towns with uncompetitive races I also addressthe important question of whether historical and geographical variation inHindu-Muslim violence is best explained using town- or state-level factors.Ashutosh Varshney, for instance, has made a good case for the primacy

of town-level factors, which he argues can constrain the actions of level ofﬁcials when it comes to riot control.40 Although, of course, bothplay a role, I show that state-level patterns of law enforcement dominate

state-38 Seymour M Hersh, “On the Nuclear Edge,” New Yorker, March 29, 1993, pp 56–73;

Devin T Hagerty, “Nuclear Deterrence in South Asia: The 1990 Indo-Pakistani Crisis,”

International Security 20, no 3 (1995–96), pp 79–114.

39 Times of India, December 8 and 9, 1992; Hindustan Times, December 11, 1992 “Damned by Faith,” Newsline (Lahore), January 1993, pp 114A–118 For information on the Bangladesh violence, see Hindustan Times, December 12, 1992.

40 Ashutosh Varshney, Ethnic Conﬂict and Civic Life: Hindus and Muslims in India (New Haven:

Yale University Press, 2002), p 10 My own view, however, developed in more detail in Chapter 2, is that state-level incentives in India are clearly dominant over local factors.

Trang 32

local factors: state law enforcement can prevent violence even in so-calledriot-prone towns and facilitate it even in towns with no previous history ofriots

If the law-and-order policies of India’s state governments are moreimportant than local-level factors in determining where Hindu-Muslimviolence takes place, the key question is obviously, What explains thesestate-level policies? In Chapters 3 and 4 I examine and test two of the majorexplanations that are usually provided to explain why some states have lowerlevels of ethnic violence than others: state capacity and governance argu-ments, and consociational arguments.41 I ﬁnd that neither differences instate capacity nor in the degree of consociational powersharing can explainthe variation we observe in states’ levels of Hindu-Muslim violence or intheir performance in preventing riots

Chapter 5 tests the main argument of the book, by examining the portance of state-level electoral incentives in explaining Hindu-Muslimviolence I show that from 1961 to 1995, higher levels of party competition

im-in the 15 major Indian states are statistically associated with lower levels ofHindu-Muslim violence I also provide qualitative evidence to show thatpoliticians do act in the way in which my model predicts and that the level ofpolitical competition for Muslim voters does have a direct effect on whether

a riot breaks out An additional question this chapter examines is whyMuslims should increasingly be the pivotal voters in Indian state politics?Why has increased political competition not placed Hindu nationalist vot-ers, rather than Muslim voters, in the pivotal position in state politics? Iargue that Muslims are especially desirable voters for Hindu politicians tocourt because of the relatively large size of their community and the rela-tively few economic and employment demands they make compared withmiddle- and lower-caste blocs of Hindu voters

If, as I argue in Chapter 5, the degree of party competition is crucial

in explaining the level of Hindu-Muslim violence in various Indian states,then it raises the question, What explains states’ different levels of partycompetition? I address this question in Chapter 6 through three case stud-ies, tracing the history of Hindu-Muslim conﬂicts and party politics in thestates of Tamil Nadu, Bihar, and Kerala I describe how, in large part because

41For the former, see Atul Kohli, Democracy and Discontent: India’s Growing Crisis of ability (Cambridge: Cambridge University Press, 1990); for the latter, see Arend Lijphart,

Govern-“The Puzzle of Indian Democracy: A Consociational Interpretation,” American Political Science Review 90, no 2 (1996), pp 258–68.

Trang 33

P1: JZP

of institutional incentives for “backward-caste” mobilization provided bythe colonial state, intra-Hindu party political competition emerged muchearlier (1920s–1930s) in the southern states of Tamil Nadu and Kerala than

in northern India I show that strong postindependence intra-Hindu petition for the Muslim vote led to governments in Kerala and Tamil Naduthat were serious about preventing and stopping Hindu-Muslim riots Thegrowing strength of similar lower- and middle-caste parties in northernIndia since the late 1980s, I predict, although it led to a short-term increase

com-in violence, will eventually lead to a similar declcom-ine com-in Hcom-indu-Muslim lence in the North

vio-In Chapter 7, I demonstrate that the electoral incentives we see at work inIndia also help account for the pattern of ethnic violence in other countries Iselect one case from each of the three great waves of democratization iden-tiﬁed by Samuel Huntington, during which multiethnic societies movedfrom uncompetitive party systems to competitive systems in a relativelyshort space of time: the “ﬁrst wave,” from 1828 to 1926, when the franchisewas extended to 50% or more of adult males in many countries in Europe,the Americas, Australia, and New Zealand; the “second wave,” after WorldWar II, when former colonies and many formerly authoritarian countries

in Latin America became democratic; and the “third wave,” which beganwith the Portuguese Revolution of 1974 and continued with democraticliberalization in Eastern Europe, the former Soviet Union, and Africa Ineach of these three cases I examine (19th-century Ireland, postindepen-dence Malaysia, and postcommunist Romania), I argue that the pattern

of ethnic violence in these countries as well as in other states such as theUnited States has been consistent with my general explanation: ethnic riotstook place where political competition was ﬁercest, and the state’s reaction

to this violence was determined by its own support base and the overalldegree of party competition in the state

The broader question this book inevitably raises is whether democraticcompetition inﬂames or reduces ethnic violence? Does the fact that electoralincentives often lead to ethnic violence mean that I agree with John StuartMill and Thomas Jefferson, both of whom at various times argued that freeinstitutions are next to impossible in multiethnic states? No In Chapter 8

I argue that, although electoral competition can foment violence, there aremany ways in which political competition as well as cleavage structures canalso be altered so that politicians have incentives to be moderate towardminorities

Trang 34

on such factors as the relative size of a town’s minority and majority tions, a town’s total population, the divisive effects caused by the presence ofrefugees from previous ethnic conﬂicts in a town, or the degree of Hindu-Muslim economic competition in an ethnically divided labor market.2 Inthe past few years, several major studies of communal violence in Indiahave also highlighted the importance of such variables as a town’s level ofinterethnic “civic engagement” or the presence or absence of “institution-alized riot systems” to explain why some towns are violent while others arenot.3

popula-This book is focused, in contrast, squarely on the state level and onpolitical incentives While town-level factors need to be taken into account,

I argue that it is even more important to understand why India’s statessometimes use force to prevent riots and at other times allow or even seem

to encourage violence Force matters because studies of riots have foundthat rioters are generally unwilling, whatever the strength of the town-level

1 Data collected by myself and Ashutosh Varshney found that 93% of deaths from 1950 to

1995 took place in towns This ﬁgure probably exaggerates the urban-rural discrepancy somewhat because riots in villages in rural areas are less likely to be reported.

2 For a review of these theories in the context of the U.S literature on race riots, see Manus

I Midlarsky, “Analyzing Diffusion and Contagion Effects: The Urban Disorders of the

1960s,” American Political Science Review 72, no 3 (1978), p 996, and Susan Olzak, The Dynamics of Ethnic Competition and Conﬂict (Stanford: Stanford University Press, 1992).

3Ashutosh Varshney, Ethnic Conﬂict and Civil Strife: Hindus and Muslims in India (New Haven: Yale University Press, 2002); Paul R Brass, Theft of an Idol: Text and Context in the Study of Collective Violence (Princeton: Princeton University Press, 1997).

Trang 35

P1: KAE

factors promoting violence, to confront armed and determined police orsoldiers who are prepared to use deadly force to stop them.4 And statesmatter, because it is India’s state governments, rather than the country’snational, municipal, or district governments, that control the local policeand paramilitary forces and decide how much force to use to prevent orstop riots at the local level Under the Indian constitution, central forcesmay only legally intervene to stop a riot if asked to do so by the local stateofﬁcials or by the state government itself This is the case even if a seriousHindu-Muslim riot breaks out only a few miles from an Indian army base,

as it did for example at Ranchi-Hatia, in Bihar, in August 1967

Despite my focus on the states and on what determines the state-levelresponse to the threat of communal riots, it is nonetheless still important totest the many influential theories about the significance of such factors asthe ethnic division of labor or the local ethnic balance in causing violence.Especially in those states where the state government is weak in ordering itsofficials to prevent violence or is openly biased, local economic, social, andpolitical factors will, I acknowledge, often be important in determining thelocation and scale of ethnic riots – in explaining why, as one journalist put

it, Bombay burned while Bhiwandi did not?5 So in this chapter I addressthe causes of this town-level variation

The Importance of Local Electoral Incentives

My main argument in this chapter is that local electoral incentives explainmuch of the variation in when and where polarizing events and commu-nal riots will break out, even when we control for towns’ previous levels

of violence and their socioeconomic attributes The idea that there is aconnection between political competition and ethnic violence is not ofcourse new, and in the previous chapter I discussed some of the broadercomparative studies that focus on political incentives to foment ethnic con-ﬂict In India, too, there has been no shortage of scholars and politicianswho have highlighted the role that electoral competition plays in precipi-tating communal violence G Ram Reddy, for example, reports that largeHindu-Muslim riots broke out in the state capital of Andhra Pradesh during

4 As Horowitz’s recent study of several hundred riots throughout the world conﬁrms, “Force seems generally to deter As police hesitation reduces inhibition in a crowd, early, determined

police action can avert what might have been a very serious riot.” Donald L Horowitz, The Deadly Ethnic Riot (Berkeley: University of California Press, 2001), pp 363–64.

5Rahul Singh, “Lessons from Bhiwandi,” Indian Express, July 18, 1993, p 8.

Trang 36

Explaining Town-Level Variation

the major municipal, state, and national elections held during the 1980s.6

Christophe Jaffrelot’s work on Madhya Pradesh has also examined the linkbetween electoral competition and riots.7Individual politicians frequentlyblame their rivals for inciting violence; Prime Minister Indira Gandhi, forexample, in a debate on large-scale riots in Gujarat in 1970, taunted theJana Sangh leader (and later prime minister) Atal Bihari Vajpayee by ask-ing him whether it was “a coincidence that when people who belong tothe RSS [Rashtriya Swayamsewak Sangh] or the Jan Sangh go somewhere,soon afterwards there is a riot close to that place?”8

But none of these individual attempts to connect electoral polarizationwith ethnic violence amounts to a general testable theory that might havesome predictive power about the specific conditions under which politi-cians have an incentive to foment violence in some constituencies and notothers In this chapter, therefore, I first build a general explanation forwhen and where specific electoral incentives will lead to violence and thentest this explanation while controlling for the main alternative explana-tions identified in other town-level analyses of violence My claim is notthat elections and electoral competition explain all town-level variation incommunal violence Given the complexity of the town-level precipitants ofviolence, putting forward a unicausal explanation of when violence breaksout would be unrealistic But I think that close electoral competition is,once we control for previous conflict and socioeconomic factors, the majorprecipitant of communal riots in contemporary India

An Electoral Incentives Model of Ethnic Riot Occurrence

A central problem facing individual politicians is how they can ensure thatvoters will identify themselves with a politician’s party and the group he

or she claims to represent, at least on polling day, rather than with other

ethnic or nonethnic groups, parties, and interests The choice of which

identity politicians choose to invoke in an election is complex and depends

on the interplay of many different factors: the extent to which existing ethnic

6 G Ram Reddy, “The Politics of Accommodation: Caste, Class and Dominance in Andhra

Pradesh,” in Francine Frankel and M S A Rao, eds., Dominance and State Power in Modern India: Decline of a Social Order, vol 1 (Delhi: Oxford University Press, 1989), pp 265–321.

7Christophe Jaffrelot, The Hindu Nationalist Movement in India (New York: Columbia

Uni-versity Press, 1996), pp 513–14.

8Lok Sabha Debates, 10th Session, 4th Series Vol XLI, No 58, May 14th 1970 (New Delhi: Lok

Sabha Secretariat), p 323.

Trang 37

P1: KAE

identities are sustained by social, religious, and economic institutions; thenumber of votes needed to win an election under any particular electoralsystem (we would not expect politicians to emphasize identities that would,even if successful in attracting their target group, result in the support oftoo small a percentage of the electorate to win the election);9 the degree

of ethnic heterogeneity within a constituency; political alliances with otherethnic parties; the strength of the party’s internal discipline; and the numberand ethnic heterogeneity of other seats in which the party is competing.10

Once politicians have decided which ethnic or nonethnic identity toinvoke, they face the challenge of how to make this identity the most polit-ically salient identity among their target voters One approach is obviously

to highlight the range of programmatic (policy) or clientelistic beneﬁts(i.e., direct transfers to speciﬁc voters) the party will deliver to the ethnicgroup once it wins the election: government jobs; subsidies to areas andeconomic sectors in which their target group is concentrated; and religiousand cultural protections.11 But, in situations where a party is dominated

by a segment of an ethnic group that enjoys a disproportionate share ofwealth, power, and government employment, promises to share the wealthwith others (whether through policy shifts or clientelistic transfers) will beviewed with skepticism by the have-nots, and with horror by those haveswho already support the party In Uttar Pradesh, for example, promises ofthe upper-caste Bharatiya Janata Party (BJP) in 1995 to introduce afﬁrma-tive action beneﬁts for “backward” Hindu castes was met with skepticism

9 For example, in India, upper-caste politicians who once formed an overwhelming majority

of the electorate have been forced to change their group appeals as the electorate has expanded from c 2% of the adult population in the 1920s to 14% after 1935 to 100%

of the adult population after 1950 See Harold Gould’s study of the town of Faizabad, which traces the changes in political appeals from the preindependence period, when only

a few thousand upper castes could vote, to the postindependence mass electorate Harold

A Gould, Grass Roots Politics in India: A Century of Political Evolution in Faizabad District

(New Delhi: Oxford & IBH, 1994), p 52.

10 For example, while it might be beneﬁcial for a politician to highlight a subethnic age such as “Presbyterian” in a by-election for a Presbyterian-dominated seat in Northern Ireland, this gain has to be set against the fact that the overall number of safe Presbyterian seats is small, and that the politician’s party may need to forge alliances with Methodists and Episcopalians in many other seats at the next election in order to defeat Catholic candidates For a general discussion of the ways in which politicians make such calculations, see

cleav-George Tsebelis, Nested Games: Rational Choice in Comparative Politics (Berkeley: University

of California Press, 1990).

11 For the distinction between programmatic and clientelist appeals, see Herbert Kitschelt,

“Linkages between Citizens and Politicians in Democratic Polities,” Comparative Political Studies 33, nos 6–7(2000), pp 845–79.

Trang 38

from backwards, who wondered about its sincerity, and succeeded in ating the BJP’s own upper-caste supporters, who forced the party to quietlywithdraw its proposals

infuri-In this situation, I argue that the most effective method for dominated ethnic parties to mobilize those target voters who are at risk

elite-of voting for the main rival parties will be to use ethnic wedge issues toincrease – albeit in the short term – the salience of ethnic issues that willfavor their party In India, both upper castes and Muslims live dispropor-tionately in urban areas in most states In Uttar Pradesh, for example, 17%

of the total state population is Muslim but Muslims account for 31% (1991census) of the state’s urban population Although no precise census ﬁguresare available for upper castes, it is generally agreed that upper castes are alsoconcentrated in urban areas, a fact reﬂected in the last caste census in 1931.Because these two groups frequently vote cohesively, they often constitutethe two main voting blocs in bipolar urban races, with the pivotal politicalposition between upper-caste- and Muslim-supported parties in towns oc-cupied by middle- and lower-caste voters The challenge for upper-castepoliticians and parties in urban areas, therefore, is how to win over thesepivotal Hindu voters

They meet this challenge by highlighting the threat posed by Muslims.Upper-caste-dominated parties can highlight anti-Muslim wedge issues –for example, Muslims’ alleged slaughter of cows, the renaming of a townwith a Muslim origin name with an “authentic Indian” (i.e., Hindu) name,12

taking a Hindu procession route through a Muslim neighborhood, or puting the status of a plot of land claimed or occupied by Muslims Thesewedge issues allow these parties to potentially rally a large proportion ofHindus (82% of the Indian population) to their side, while entailing noeconomic cost for the party’s existing upper-caste supporters In Indian

dis-terms, the upper castes are ﬁghting Mandal – the name of a commission

that in 1980 recommended large-scale afﬁrmative action programs for the

backward castes – with Mandir (a Hindu temple).13

The particular form of antiminority mobilization used depends on boththe identity politicians wish to make salient and the fact that the Indian state,

12The BJP proposed in 1990 and 2001 that Ahmedabad be renamed “Karnavati.” Hindu,

June 11, 2001 Similar proposals have been made to rename Allahabad “Prayag.”

13 The incentives for Muslim candidates to polarize the vote exist theoretically but not often

in practice because Muslims are 40% or more of the population in only 11 of the 219 largest

cities in the country, and constitute a majority in only 6 R Ramachandran, Urbanization and Urban Systems in India (Delhi: Oxford University Press, 1992), p 177.

Trang 39

P1: KAE

like other states, institutionally privileges some forms of mobilization – and,

in particular, “traditional” religious ceremonies and processions – over ers.14A favorite strategy of Hindu party leaders who calculate that they willgain electorally from polarization around a Hindu identity is to organizeunusually large religious processions that take new routes through minorityneighborhoods, to hoist the national flag over a disputed site, or to spon-sor processions to celebrate national anniversaries These tactics make itvery difficult for the local administration to ban the event, for who couldpossibly object to the performance of a religious obligation, the raising ofthe national flag, or the celebration of a national day?15But the organizers,once permission has been granted, can easily introduce symbols and speechinto these events that is likely to provoke the other community.16

oth-If members of the other ethnic group gather to watch the event anddefend their neighborhood or community symbols, this countermobiliza-tion can then be portrayed as an illegitimate provocation by the minorities

on the part of the organizers Defensive countermobilization by ties also greatly increases the probability of ethnic violence because, whencrowds face each other, the power of individuals to influence their group’sactions – whether that individual is a political organizer who wants to in-cite violence deliberately or a nervous youth intimidated by members ofthe other community – becomes magnified enormously If one demonstra-tor throws a stone, it is interpreted as “the crowd” throwing stones: if oneMuslim or Catholic or Jew fires a shot, it is interpreted as “the Muslims”

minori-or “the Catholics” minori-or “the Jews” shooting The instant this kind of violent

14 The principle that government should be “neutral” toward religions and allow, to the maximum extent possible, each religion to carry out processions and ceremonies began to

be introduced in the 1830s, and became a cornerstone of post 1857-government policy, despite the conﬂict and political mobilization around religious identities it has caused For

an examination of this policy shift and how it was related to 19th-century riots, see C A.

Bayly, Ruler, Townsmen and Bazaars: North Indian Society in the Age of British Expansion, 1770–1870 (Cambridge: Cambridge University Press, 1988), pp 335–38, and Katherine

Prior, “Making History: The State’s Intervention in Urban Religious Disputes in the

North-Western Provinces in the Early Nineteenth Century,” Modern Asian Studies 27, no 1 (1993),

pp 200–2.

15 Administrations in such circumstances must always tread a fine line between preventive action sufficient to prevent riots and preventive action that is so heavy-handed (mass arrests, beating of religious figures, etc.) that it begins to alienate large swaths of the majority community.

16 For a ﬁne analysis of processions as a form of mobilization, see Christophe Jaffrelot, “The Politics of Processions and Hindu-Muslim Riots,” in Amrita Basu and Atul Kohli, eds.,

Community Conﬂicts and the State in India (Delhi: Oxford University Press, 1998), pp 58–

92.

Trang 40

action occurs, a crowd member’s identity becomes completely and untarily subsumed to that of his ethnic group As James Rule points out,

invol-“The behavior of many, perhaps most individuals in the crowds may nothave changed, yet the social construction of their actions may move themfrom the non-rioter category into that of rioters.”17

The minority countermobilization or ethnic violence that results fromthis kind of electoral mobilization will not, of course, be sufﬁcient to scareall the Hindu swing voters into rallying behind the most pro-Hindu party.Many voters, after all, will have ﬁrm political allegiances to particular ideo-logical or ethnic political parties And some voters will have a greater degree

of bias toward minorities than others But to win an election it is not essary to appeal to every voter but only to pivotal swing voters, especiallythose undecided voters who are uninformed, unlikely to vote (unless scaredinto doing so), and most likely for whatever reason to fear the consequences

nec-of not taking a strong defensive posture toward members nec-of the other ethnicgroup.18 In the southern United States, for example, James Glaser inter-viewed campaign managers who had a clear sense that ethnic wedge issueswould appeal more to some groups among the white electorate than oth-ers For example, one campaign manager told him that rural white voterswere normally Democratic but that racial issues could swing them to theRepublicans.19

Organizing processions and other types of mobilization designed tohighlight ethnic cleavages requires scarce resources: time, effort, andmoney Therefore we should not expect divisive ethnic mobilization to takeplace at all times or in every seat in which ethnic parties compete First,

it seems likely that polarizing events will occur disproportionately beforeelections as politicians try use inﬂammatory issues to solidify their ownethnic community’s support or to intimidate their ethnic opponents.20InKenya, for example, Daniel Arap Moi was accused of fomenting intertribal

17See James B Rule, Theories of Civil Violence (Berkeley: University of California Press, 1988),

p 47.

18William H Riker, The Strategy of Rhetoric: Campaigning for the American Constitution (New

Haven: Yale University Press, 1996), pp 50–51.

19James M Glaser, Race, Campaign Politics and the Realignment in the South (New Haven: Yale

University Press, 1996), p 69.

20 If, on the other hand, democratic politics are not well institutionalized within a country (in the ﬁrst election after a period of authoritarian rule, for instance) we would expect to see a different pattern of ethnic violence, as losers challenge the legitimacy of the electoral process itself This happened in the Congo, for example, where more than 2,000 died in

ethnic violence following the 1993 election Agence France Presse, January 13, 1995.

Định dạng
Số trang	311
Dung lượng	1,64 MB